This study explored the role of question types and prior knowledge in vicarious learning with an intelligent tutoring system. In experiment 1, the participants were assigned to three conditions (deep questions, shallow questions, control), the results showed that participants in the deep questions condition had higher retention test scores than those in the shallow questions condition and those in the control condition. In experiment 2, we conducted a 2 (prior knowledge: high, low) × 3 (question types: deep question, shallow question, control) between-subjects experimental design, and participants in the deep question condition again had higher retention test scores than the other conditions. Also, participants in the shallow question condition had higher retention test scores than those in the control condition. Additionally, no interaction between prior knowledge and question types was observed. In conclusion, deep questions can promote learning of experimental design, which replicates prior research results on deep level questions.